test(delegate): tidy #121572 follow-ups

elide() already returns short text unchanged, so the eval runner's length
guard was a second copy of the same check. The timeout-diagnostic test
read its log as utf-8-sig although the writer never emits a BOM, and it
interrupted a stub child that never ran.

evals/session_search_schema/runner.py

@@ -177,8 +177,7 @@ def run_one(client, model, arm_name, arm_mod, task_id, prompt, oracle,
                 out, was_err = exec_tool(arm_mod, args, main_db_path)
             if was_err:
                 bad_calls += 1
-            if len(out) > 30000:
-                out = elide(out, 30000)
+            out = elide(out, 30000)
             messages.append(
                 {"role": "tool", "tool_call_id": tc.id, "content": out})
     return {

tests/tools/test_delegate_subagent_timeout_diagnostic.py

@@ -126,8 +126,7 @@ class TestDumpSubagentTimeoutDiagnostic:
             worker_thread=None,
             goal="g" * 1200,
         )
-        child.interrupt()
-        text = Path(path).read_text(encoding="utf-8-sig")
+        text = Path(path).read_text(encoding="utf-8")
         assert _COMPRESSION_MARKER_RE.search(text)
         assert "g" * 100 in text
         assert "...[truncated]" not in text